AI clustering around human median while economists scatter widely

Give 146 economist teams the same dataset and you get wildly different answers. Rerun that setup with agentic AI, and Claude Code and Codex land near the human median instead of collapsing or spiking to outliers.

The shift is variance. The AI runs cluster tightly with no extreme outputs, while the human sample spreads wide. The paper frames that consistency as useful for scaling structured empirical work: you get middle-of-the-pack results, but you get them reliably.